awesome-ai-agent-papers: A curated collection of AI agent research papers released in 2026, covering agent engineering, memory, evaluation, workflows, and autonomous systems

Summary

VoltAgent's GitHub repository is a carefully curated, weekly updated directory of over 400 arXiv papers from 2026. It covers AI agent subfields such as multi-agent coordination, memory/RAG, evaluation, tools, and security, providing a one-to-two-sentence summary for each paper.

Why it matters

The AI agent literature on arXiv is growing rapidly and remains fragmented. A single classified and actively maintained list reduces discovery costs for engineers and researchers who need to track what has actually been released versus what is still in preprint stage. It also highlights the subfields attracting the most research attention in 2026 (approximately 100 papers on tools and approximately 80 on evaluation).

Highlights

• Approximately 400 papers from arXiv, updated weekly and categorized into five groups
• Notable entries include AutoNumerics (PDE solver generation), CORAL (self-evolving multi-agent systems, performing 3–10 times better than baselines), and AgenticPay (110+ task negotiation benchmark)
• The repository has 1.8k stars but 0 open issues and 0 pull requests, indicating that consumption is progressing much faster than community contribution

Critical take

The 'curated' framing contradicts a list that swallows nearly all agent-related arXiv submissions into five broad categories. Additionally, four paid sponsorship placements in the README raise questions about editorial independence. Zero open issues and zero pull requests suggest that this compilation is the work of a single maintainer rather than a community-reviewed resource.